Anthropic Identity Verification: What the June 2026 Notices Actually Said
Anthropic's June 2026 notices said it may request age or identity confirmation in some consumer-account cases; they did not announce universal passport checks.
Contents
Anthropic Identity Verification: What the June 2026 Notices Actually Said
This is a historical review of notices published in June 2026, not a statement of what every Claude user must do today. If the task only needs permitted model API access, open BetterToken’s current Claude Code guide, configure your own account and API Key, run one small read-only request, note its start time, and match the response with the model, status, and usage entry created in Dashboard. This API route does not replace Claude.ai consumer features or bypass Anthropic’s safety, identity, location, or account rules.
What the June 8 notice said
Anthropic’s Privacy Center announced changes effective July 8, 2026 for consumer Claude Free, Pro, and Max accounts. It said Anthropic may ask a user to confirm age or identity, and added explanations of data used for multi-step tasks, connected apps, and voluntary studies. The notice explicitly excluded Team, Enterprise, the Claude Developer Platform, and services governed by other agreements from that particular consumer-policy update.
What Anthropic later confirmed
On June 17, Anthropic’s Help Center said identity verification was being rolled out for a few use cases. A prompt could appear for certain capabilities, routine platform-integrity checks, or other safety and compliance measures. That wording does not establish a universal requirement, a per-session check, or a fixed trigger list.
ID, selfie, and Persona: the documented boundary
The Help Center names Persona Identities as Anthropic’s verification partner. A person asked to complete this identity-verification flow may need a valid physical government photo ID and may be asked for a live selfie. Those are requirements for the flow when it is presented, not proof that every user will receive it. Anthropic says it is the controller of verification data; Persona processes it under Anthropic’s instructions. Persona collects and holds the ID and selfie, while Anthropic may access verification records when needed, for example for an appeal, but says it does not copy or store those images itself. The data is not used to train models.
What remained unknown in June
The policy update and Help Center did not publish a universal schedule, a fixed trigger list, or universal consequences for refusing or failing to finish verification. Do not invent those consequences or infer another account’s conditions from a screenshot, email, or social-media discussion. Age assurance is a separate process with its own provider and methods; it should not be conflated with Persona identity verification.
Practical response for a work account
- No prompt: do not upload documents proactively. Open the official Privacy Center and Help Center directly, update recovery information, and map the account, provider, API Key, and billing dependency for each critical workflow.
- Prompt in a signed-in Claude session: open
claude.aidirectly instead of following an email link, return to the signed-in prompt, and verify that it links to the official privacy or help pages. Continue only to a partner flow launched from the official product; otherwise stop and contact support. - Restricted account: record the exact message, time, and plan without copying documents, API Keys, or other secrets. Use Anthropic’s official support route. Do not misrepresent identity, region, or payment information.
An independent API route is not a compliance workaround
BetterToken provides an API service with a user’s own BetterToken account and API Key. Use this route only when the workflow needs model API access; it does not provide Claude.ai accounts, shared accounts, consumer-only features, or an exemption from Anthropic’s policies. For a small verification request, note the start time first. PASS means the response completes without 401 or 403 and a new Dashboard entry after that time shows the expected model, status, and usage. Otherwise treat the test as failed, compare the configuration with the current guide, and use the appropriate support route.
